Acquiring sharply-focused images of moving people or objects is a fundamental and challenging problem in several surveillance applications, particularly iris-based biometrics and face recognition.
Iris recognition is a method of biometric authentication that utilizes pattern recognition techniques based on high-resolution images of the irises of an individual's eyes. Iris recognition relies on camera technology with subtle infrared illumination reducing specular reflection from the convex cornea to create images of the detail-rich, intricate structures of the iris. Converted into digital templates, these images provide mathematical representations of the iris that yield unambiguous positive identification of an individual.
Iris recognition has been recently recognized and gained much attention due to its high reliability in identifying humans. Its suitability as an exceptionally accurate biometric derives from its extremely data-rich physical structure, genetic independence (no two eyes are the same even for twins), stability over time, and non-contact means (a feature important for non-cooperative subjects).
Facial recognition typically involves the use of a computer application for automatically identifying or verifying a person from a digital image or a video frame from a video source. One of the ways to accomplish this is by comparing selected facial features from the image and a facial database.
One type of device that has been found in use in biometric identification and authentication is the flutter shutter camera, based on flutter shutter technology. Note that a non-limiting example of a flutter shutter camera and flutter shutter technology in general is disclosed in U.S. Patent Application Publication Serial No. US2007/0258707A1, entitled “Method and Apparatus for Deblurring Images,” which published to Ramesh Raskar on Nov. 8, 2007, and is incorporated herein by reference. Another non-limiting example of a flutter shutter camera and flutter shutter technology is disclosed in U.S. Patent Application Publication Serial No. US2007/0258706A1, entitled “Method for Deblurring Images Using Optimized Temporal Coding Patterns,” which published to Ramesh Raskar, et al on Nov. 8, 2007, and is incorporated herein by reference.
Iris-based biometric identification thus has great potential to enhance controlled access and even surveillance of high security areas. Much of this potential can only be realized by acquiring iris images from non-cooperative subjects. Like other applications that exploit visual information, however, iris-based biometrics is limited by the ability to acquire high-quality images in certain situations. One situation that is particularly challenging is the acquisition of sharply-focused iris images from moving subjects. Given a modest amount of light, as is common indoors, relatively long exposures are necessary at even the widest aperture setting. For moving subjects, this produces motion-blurred images which lack much of the crucial high-frequency information needed to perform iris matching.
For an application like iris recognition, wherein fine scale features are essential to proper classification, the use of a traditional shutter imposes some fundamental limits on the extent of motion blur that can be tolerated. Motion blur, as through a traditional shutter, is equivalent to convolution of a sharply-focused image with a box filter. Motion-blurred images of this type lack information regarding the object at a number of spatial frequencies. This lack of information is irreversible and no post processing can recover it from the image. Methods that attempt to deblur the image will severely amplify sensor noise, hallucinate content, or both. Though it may be useful in improving subjective image quality, hallucinating image content is counter-productive for forensic applications, and amplifying noise complicates iris matching.
To avoid this loss of information during image capture, some prior art approaches have advocated the use of a fluttering shutter and demonstrated the ability to recover high-quality images despite blur from moving objects. During exposure, the camera's shutter flutters between open and closed while exposure is accumulated on the sensor. This produces an image with coded blur which, unlike traditional blur, conveys information about the subject at all spatial frequencies. Given a suitably designed processing method that is based on the shutter's fluttering pattern, deblurring recovers an image with low levels of noise while avoiding reconstruction artifacts.
In moderate and low lighting conditions, however, the need to use long exposure times causes images of moving subjects to be significantly degraded by blur. This blur will destroy fine details of, for example, the subject's iris to an extent that the iris image will be useless for identification. It is therefore necessary to design systems that are capable of acquiring sharp images of subjects despite low lighting conditions and subject motion.